Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving

 
CONTINUE READING
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
electronics
Article
Deep Feature-Level Sensor Fusion Using Skip Connections for
Real-Time Object Detection in Autonomous Driving
Vijay John * and Seiichi Mita

                                           Research Center for Smart Vehicles, Toyota Technological Institute, Nagoya 468-8511, Japan; smita@toyota-ti.ac.jp
                                           * Correspondence: vijayjohn@toyota-ti.ac.jp

                                           Abstract: Object detection is an important perception task in autonomous driving and advanced
                                           driver assistance systems. The visible camera is widely used for perception, but its performance is
                                           limited by illumination and environmental variations. For robust vision-based perception, we pro-
                                           pose a deep learning framework for effective sensor fusion of the visible camera with complementary
                                           sensors. A feature-level sensor fusion technique, using skip connection, is proposed for the sensor
                                           fusion of the visible camera with the millimeter-wave radar and the thermal camera. The two net-
                                           works are called the RV-Net and the TV-Net, respectively. These networks have two input branches
                                           and one output branch. The input branches contain separate branches for the individual sensor
                                           feature extraction, which are then fused in the output perception branch using skip connections. The
                                           RVNet and the TVNet simultaneously perform sensor-specific feature extraction, feature-level fusion
                                           and object detection within an end-to-end framework. The proposed networks are validated with
                                           baseline algorithms on public datasets. The results obtained show that the feature-level sensor fusion
                                           is better than baseline early and late fusion frameworks.

                                 Keywords: deep sensor fusion; intelligent vehicles
         

Citation: John, V.; Mita, S. Deep
Feature-Level Sensor Fusion Using
Skip Connections for Real-Time
Object Detection in Autonomous             1. Introduction and Literature Review
Driving. Electronics 2021, 10, 424.
                                                 Environment perception tasks, such as object detection, is an important research topic
https://doi.org/10.3390/
electronics10040424
                                           in autonomous driving and advanced driving assistant system [1,2]. The results obtained
                                           from perception are used by the autonomous driving systems for navigation and control.
Academic Editors: Dong Seog Han,           A similar application of the perception task is found in wireless sensor networks [3–5], where
Kalyana C. Veluvolu and Takeo Fujii        the perception obtained from the sensor nodes are sent to actuator nodes for intelligent
Received: 13 January 2021                  traffic monitoring.
Accepted: 8 February 2021                        Research Problem: For object detection, typically, an array of different sensors such as
Published: 9 February 2021                 visible camera, stereo camera, thermal camera and radar are used. Among these different
                                           sensors, the visible camera is widely used. However, the visible camera-based object
Publisher’s Note: MDPI stays neu-          detection is affected by environmental and illumination variations.
tral with regard to jurisdictional clai-         In this paper, we address the limitation associated with the visible camera-based
ms in published maps and institutio-       object detection by proposing novel sensor fusion frameworks using deep learning and
nal affiliations.                          complementary sensors. The proposed deep learning frameworks simultaneously perform
                                           sensor-specific feature extraction, feature-level sensor fusion and object detection in an
                                           end-to-end manner. Two deep learning frameworks called the RVNet for radar-visible
Copyright: c 2021 by the authors.
                                           camera-based object detection and the TVNet for thermal-visible camera-based object
Licensee MDPI, Basel, Switzerland.
                                           detection are presented in this paper.
This article is an open access article
                                                 The radar and thermal camera are complementary sensors used by the RVNet and
distributed under the terms and con-       TVNet for object detection. A comparison between the visible camera and the complemen-
ditions of the Creative Commons At-        tary sensors are illustrated in Tables 1 and 2 and Figures 1 and 2. As illustrated, the sensor
tribution (CC BY) license (https://        pairs have complementary characteristics, and their fusion will enhance the perception
creativecommons.org/licenses/by/           robustness in varying conditions. We next review the literature before highlighting the
4.0/).                                     contribution of our work.

Electronics 2021, 10, 424. https://doi.org/10.3390/electronics10040424                                       https://www.mdpi.com/journal/electronics
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                                      2 of 12

                            Figure 1. Scenes from the Nuscenes dataset, where radar reflection are obtained in varying conditions.
                            The camera obtains the object boundaries in the same conditions.

                               Day: Sun lare avoided in thermal camera            Day: Pedestrian observed in both cameras

                              Night: Pedestrian observed in thermal camera         Night: Car observed in headlight glare in
                                                                                               thermal camera

                             Night: Pedestrian and car observed in thermal            Dusk: Car observed in thermal camera
                            Figure 2. Scenes from the FLIR dataset, where thermal camera images are compared with visible
                            camera images.

                                  Literature Review: In case of the sensor fusion of the radar and visible camera, the
                            sensor fusion techniques are classified as early fusion-based obstacle detection [6–8], late
                            fusion-based obstacle detection [9–11] and feature fusion-based obstacle detection [12,13].
                            In early fusion, first, a set of candidate obstacles detected using the radar are iden-
                            tified in the visible camera. Subsequently, camera-based algorithms [14,15] are used
                            to prune the candidates and identify the final set of obstacles. Bombini et al. [6] and
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                             3 of 12

                            Sugimoto et al. [7] adopt such a strategy, where the radar object candidates are transferred
                            to the camera coordinate for vision-based object detection. In late fusion-based obstacle
                            detection, radar-based obstacle detection and camera-based obstacle detection are per-
                            formed independently. The individual detection results are heuristically combined and
                            the final set of obstacles are detected [9,10]. In feature fusion-based object detection, the
                            radar and vision features are fused in a deep learning-based framework [12,13,16], where
                            simultaneous sensor fusion and obstacle detection is performed. Chadwick et al. [12]
                            propose a deep learning framework based on the SSD to detect obstacles in the visible
                            image using both radar and visible camera features.
                                  Compared to the visible-radar literature, the RVNet framework improves the state-
                            of-the-art detection accuracy with real-time computational complexity. Additionally, the
                            RVNet proposes a novel radar feature map called the sparse radar image for object detec-
                            tion.
                                  Unlike radar and vision-based obstacle detection, the focus of the thermal and visible
                            camera-based obstacle detection is different. Here, researchers focus on generating an
                            enhanced multimodal visible-thermal image by fusion, instead of enhancing the object
                            detection algorithm. Researchers report improved detection accuracy by enhancing the
                            input to the object detection framework, without any change to the detection framework.
                            The generated multimodal image contains the visible camera’s rich appearance information
                            as well as the thermal camera’s heat signature information [17–22]. The multi-modal images
                            are obtained using multi-resolution schemes [17–20], local neighborhood statistics [23] and
                            learning methods [24–26]. Shah et al. [17] generate the multi-modal image using wavelet
                            analysis and contourlet transform. The pixel value in the fused image is obtained by the
                            weighted average of the thermal and visible pixels. The weights are calculated using the
                            significance of the source pixels. For example, Shah et al. [23] use the local neighborhood
                            eigenvalue statistics to calculate the significance of the source pixels. Liu et al. [18] propose
                            a multi-resolution fusion scheme to generate the multimodal image. The visual image is
                            enhanced using contextual information derived from the thermal images. Zhang et al. [21]
                            generate the image using quadtree decomposition and Bezier interpolation to obtain an
                            enhanced thermal image with distinct thermal-based features. This enhanced image is
                            added to the visible image to generate the multimodal image. John et al. [22] generate a
                            multi-modal image using pixel-level concatenation, and perform pedestrian detection.
                                  In recent years, with the advent of deep learning framework, researchers have also
                            used deep learning for this task. Shopovska et al. [24] generate the multimodal image
                            using deep learning, where the fused image is visually similar to a color image, but
                            contains important features in the pedestrian regions. In the work by Ma et al. [25],
                            the FusionGAN, a GAN framework, is used to generate the multimodal image. An
                            extension of the FusionGAN is proposed by Zhao et al. [26]. The generated multimodal
                            images are given to object detection algorithms for robust perception [22].
                                  Compared to the existing thermal-visible camera literature, the TVNet framework
                            does not focus on generating the multimodal image. Instead the TVNet is a sensor fusion
                            framework that simultaneously performs sensor-specific feature extraction, feature-level
                            sensor fusion and also object detection within an end-to-end framework. The proposed
                            formulation is shown to improve the detection accuracy with improved robustness.
                                  Proposed Framework: Both the RVNet and the TVNet have similar deep learning
                            architectures. Both the networks contain two input branches and one output branch.
                            The input branches contain separate branches for the individual sensor feature extraction.
                            The input branches extract the sensor-specific features. These features are transferred to
                            the output branch using the skip connection, where the feature-level fusion and object
                            detection is performed. The output branch performs single shot object detection and is
                            based on the YOLO network [27]. An overview of the architecture is shown in Figure 3.
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                                 4 of 12

                                   Visible Camera Feature
                                     Extraction Branch
                                                                        Output Fusion
                                                                         and Obstacle
                                Complementary Sensor Feature           Detection Branch
                                     Extraction Branch

                            Figure 3. An overview of the skip connection-based fusion and obstacle detection architecture.

                                The networks are validated on the Nuscenes [28] and FLIR public dataset [29] and
                            compared with state-of-the-art baseline algorithms. The experimental results show that the
                            proposed feature-level fusion is better than baseline sensor fusion strategies.
                                Key Contributions: To the best of our knowledge, the main contributions of the
                            proposed framework are as follows:
                            •       The TVNet is a novel end-to-end deep learning framework which simultaneously
                                    extracts sensor specific features, performs feature-level fusion and object detection for
                                    the thermal and visible cameras.
                            •       Comparative experimental analysis of early, late and feature-level fusion of visible
                                    and thermal camera for obstacle detection.

                            Table 1. Comparative study of the properties of the radar and visible camera.

                                              Sensor                 Milliwave Radar                  Visible Camera
                                                                       Not affected by                  Affected by
                                              Weather
                                                                   rain, snow and fog [30]          rain, snow and fog
                                         Illumination                    Not affected                     Affected
                                         Data density                Sparse appearance              Dense appearance
                                        Object boundary                      No                             Yes
                                         Object Speed                        Yes                            No

                            Table 2. Comparative study of the properties of the thermal and visible camera.

                                              Sensor                 Thermal Camera                   Visible Camera
                                                                       Not affected by                  Affected by
                                             Weather
                                                                   rain, snow and fog [30]          rain, snow and fog
                                                                      Not affected by                   Affected by
                                          Illumination                low-light and                    low-light and
                                                                       low-visibility                  low-visibility
                                                                       Not affected by                  Affected by
                                            Lens flare               direct sunlight and            direct sunlight and
                                                                          headlight                      headlight

                                 The remainder of the paper is structured as follows. The skip connection networks
                            are presented in Section 2. The experimental analysis is presented in Section 3. Finally,
                            the paper is concluded in Section 4.

                            2. Deep Learning Framework
                                 Our proposed deep learning frameworks simultaneously performs sensor specific feature
                            extraction, feature-level fusion and object detection in an end-to-end learning framework.
                                 The vision-based fusion with radar is performed by the RV-Net, while the fusion
                            with the thermal camera is performed by the TVNet. Both the RV-Net and the TV-Net
                            contains two input branches, and one output branch. The independent input branches
                            extract sensor specific features relevant for object detection. The features extracted are
                            transferred to the output branch for feature fusion and obstacle detection using the skip
                            connections [31]. The output branch is a single shot object detection head, based on the tiny
                            YOLO network [27] detecting all the objects in the driving environment within a binary
                            classification framework.
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                                                                                       5 of 12

                            2.1. RVNet: Radar and Visible Camera Fusion and Object Detection
                                 The RVNet is a deep learning framework with two input modules and one output
                            module as shown in Figure 4. The two independent input branches extract visible camera
                            I features and millimeter wave radar S features, respectively. The radar specific features
                            are extracted from the radar input. The radar input to the RVNet, is not the raw radar data,
                            but rather a reformulated sparse radar image.

                               Visible Camera Feature Extraction Branch
                                                                                                                                     Conv2D(256,1)
                                                                                  Image
                                                   Darknet Tiny Feature
                                                                                 Feature                        Concatenaon
                                                        Extracon                                                                     Conv2D(384,3)
                                                                                  Maps                          [x_fusion_up-
                                                       (Tiny Yolov3)                                             sample, x_8]
                                                                                 [x,x_8]
                                                                                                                                      Conv2D(128,3)
                                Image
                              (416,416,3)
                                              Tiny YOLOv3 feature extraction
                                                                                                                Upsampling2D           Yolo Output
                                                                                                                    (2,2)                  Conv
                                                                                                                                                            Smaller obstacles
                                            Sparse radar feature extraction branch                                                    Conv2D(30,1)
                                                                                                                Conv2D(128,1)                                detecon and
                                                                                                                                          Output
                                                                                                                                                              classicaon

                                                                                            Skip connections
                                                                                                                                         Reshape
                                                      Conv2D (64,3)        Conv2D (128,3)

                                                                               Maxpooling                      Concatena on           Fused Feature
                                                     Conv2D (64,3)
                                                                                                                [x, x_radar]              Maps
                                                                           Conv2D (128,3)                                               [x_fusion]
                                                       Maxpooling
                              Sparse Radar
                                 Image                                         Maxpooling                      Conv2D(128,1)
                               (416,416,3)           Conv2D (256,3)

                                                                           Conv2D (128,3)                                              Yolo Output
                                                       Maxpooling                                                                         Conv
                                                                                                               Conv2D(192,3)
                                                                                                                                      Conv2D(30,1)            Big obstacles
                                                     Conv2D (256,3)
                                                                                Radar                                                   Output                detecon and
                                                                            Feature Maps                       Conv2D(64,3)             Reshape               classicaon
                                                       Maxpooling
                                                                              [x_radar]
                                                                                                                                  Fusion across two output branches

                                     Radar Feature Extraction Branch                                                            Output Fusion Branch

                            Figure 4. An overview of the RVNet architecture.

                                 Sparse Radar Image: The sparse radar image is a highly sparse image with pixel-
                            level radar information, corresponding to depth, lateral velocity and longitudinal velocity.
                            To generate the sparse radar image, the raw radar points are first transformed from the
                            radar coordinate system to the camera’s coordinate system using the extrinsic calibration
                            parameters. Subsequently, using the camera’s intrinsic calibration parameters, the trans-
                            formed raw radar points are formulated as the sparse radar image S. The sparse radar
                            image’s size is same as the image (416 × 416), and contains 3-channels corresponding to
                            the radar features.
                                 Input Branches: The visible camera feature extraction branch is based on the pre-
                            trained Darknet model used in the tiny Yolo3 model [27]. The Darknet model is pre-trained
                            on the Pascal VOC dataset [32].
                                 The radar feature extraction branch is formulated to extract features from the highly
                            sparse radar image S. Multiple 2D convolution filters with a single stride are used to
                            account for every pixel-level radar feature. To reduce the dimensionality of the feature
                            maps, maxpooling 2D is used.
                                 Output Branch: The features extracted from the sparse radar image and the visible
                            camera are transferred to the output branch using skip connections. The individual feature
                            maps are concatenated with each other at various levels in the output branch. The output
                            branch based on the tiny Yolov3 model contains two sub-branches. The first sub-branch
                            detects small and medium-sized obstacles, and the second sub-branch detects big obstacles.
                            The output branch uses multiple 2D convolution filters and reshapes layers.

                            2.2. TVNet: Thermal and Visible Camera Fusion and Object Detection
                                Similar to the RVNet, the TVNet also contains two input modules and one output
                            module as shown in Figure 5. The two independent input branches extract visible camera
                            image, I, features and thermal camera image, T, features, respectively.
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                             6 of 12

                                  As a pre-processing step, the thermal image is registered with the visible camera
                            coordinate system. The thermal camera image is single channeled (416 × 416) image,
                            which is the same size as the visible image.
                                  Input Branches: Similar to the RVNet, the visible camera feature extraction branch in
                            the TVNet is also based on the Darknet model used in the tiny Yolo3 model [27] to use the
                            pre-trained weights of the model previously trained on the Pascal VOC dataset [32].
                                  In case of the thermal camera feature extraction branch, the layer architecture is similar
                            to the Darknet model used in the tiny Yolo3 model. However, owing to the input thermal
                            camera image being single channeled, the pre-trained weights cannot be used for the
                            thermal camera feature extraction branch.
                                  Output Branch: The features extracted from the thermal and visible images are trans-
                            ferred to the output branch using skip connections. The architecture of the output branch is
                            similar to the RVNet’s output branch, including the concatenation of the individual feature
                            maps. Similar to the RVNet, the output layer of the TVNet is also trained with the YOLO
                            loss.

                                    Figure 5. An overview of the TVNet architecture.

                            2.3. Training
                                  Both the RVNet and the TVNet are trained as binary obstacle detectors on public
                            datasets. The RVNet is trained with the nuscenes dataset, and the vehicles, motorcycles,
                            cycles, pedestrians, movable objects and debris classes are considered to be obstacles [28].
                            On the other hand, the TVNet is trained with the FLIR dataset, where car, pedestrians,
                            dogs and bicycles are considered to be obstacles [29].
                                  In both the RVNet and TVNet, the pretrained weights of the feature extraction layers
                            of the tiny YOLOv3 network trained on the Pascal VOC dataset [32] are used to initialize
                            the weights of the image feature extraction branch.
                                  On the other hand, the radar and thermal feature extraction and the output branches
                            are initialized with random weights and trained without any fine-tuning. The networks
                            are trained with an Adam optimizer and learning rate of 0.001.
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                                                       7 of 12

                            3. Experimental Results and Discussion
                                 Dataset: The RVNet was validated on the nuscenes dataset and the TVNet was vali-
                            dated on the FLIR dataset. The nuscenes dataset contains 3200 training samples and 1000
                            testing samples. The dataset contains scenes from daytime, nighttime and rainy weather
                            (Figure 1).
                                 The FLIR dataset contains a day subset and a night subset. The day subset con-
                            tains 3714 training and 1438 test samples. The night subset contains 3670 training and
                            1482 testing samples. The dataset contains scenes from daytime and nighttime (Figure 2).
                                 RVNet Baseline Algorithms: Two algorithms based on the tiny YOLOv3 network [27]
                            are used as the baseline algorithms. The first baseline algorithm is the original Tiny Yolov3
                            network. which is trained on the visible camera image alone. However, unlike the original
                            network, the baseline tiny Yolov3 networks is trained as a binary classifier.
                                 The second baseline algorithm is a late fusion-based object detection network, where
                            the results obtained from independent radar and vision object detection pipelines are
                            combined in the final stage. The visible camera pipeline is a tiny Yolov3 network modified
                            as a binary classifier. The radar pipeline is a rule-based classifier, where radar’s velocity
                            is used to identify the centroids of obstacles. In the absence of obstacle bounding box
                            information, a pre-defined bounding box is generated around the detected centroids. The
                            difference between the baseline and the RVNet is summarized in Figure 6.

                                Algo       Sensor     Feature Extraction       Feature          Object                       Notes
                                           Fusion                              Fusion          Detection
                             TinyYolov3     No      Visible camera feature       No      Visible camera features   Visible camera-based
                             (Visible)                                                             used            object detection
                             Late Fusion    Yes     Visible camera and radar     No        Visible camera and      Detection results from two
                                                    specific features                      radar features used     independent TinyYolov3
                                                                                             independently         combined in final step
                             RVNet          Yes     Visible camera and           Yes      Visible camera and       Simultaneous feature
                                                    radar specific features               radar features used      extraction, feature fusion
                                                                                                 jointly           and joint object detection

                            Figure 6. RVNet compared with the baseline algorithms. Contributions to literature highlighted.

                                 TVNet Baseline Algorithms: Multiple algorithms based on the tiny YOLOv3 net-
                            work [27] are used as the baseline algorithms.
                                 The first and second baseline algorithms are the original Tiny Yolov3 network trained
                            on the visible camera image and thermal camera image, respectively.
                                 The third baseline algorithm is an early fusion-based object detection framework,
                            where a multimodal thermal-visible camera image is given as input to the tiny YOLOv3
                            architecture. The multimodal image is generated by the concatenation of registered thermal
                            and visible camera images.
                                 The final baseline algorithm is a late fusion-based object detection framework, where
                            the results obtained from independent thermal and vision object detection pipelines are
                            combined in the final stage. The independent pipelines are the tiny YOLOv3 networks,
                            and the results obtained from the pipelines are combined. A non-maximum suppression is
                            performed on the combined results to obtain the final result. The difference between the
                            baseline and the TVNet is summarized in Figure 7.
                                 Algorithm Parameters: The performance of the networks are reported using the
                            Average Precision (AP) with IOU (intersection over threshold) of 0.5.
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                                                        8 of 12

                                Algo        Sensor     Feature Extraction       Feature           Object                      Notes
                                            Fusion                              Fusion           Detection
                             TinyYolov3      No      Visible camera feature       No      Visible camera features   Visible camera-based
                             (Visible)                                                              used            object detection
                             TinyYolov3      No      Thermal camera feature       No         Thermal camera         Thermal camera-based
                             (Thermal)                                                        features used         object detection
                             Early           Yes     Multimodal features          Yes      Multimodal features      Concatenated multimodal
                             Fusion                                                              used               visible-thermal image used
                                                                                                                    for object detection
                             Late Fusion     Yes     Visible and thermal          No        Visible and thermal     Detection results from two
                                                     camera specific features              camera features used     independent TinyYolov3
                                                                                              independently         combined in final step
                             TVNet           Yes     Visible and thermal          Yes      Visible and thermal      Simultaneous feature
                                                     camera specific features             camera features used      extraction, feature fusion
                                                                                                  jointly           and joint object detection

                            Figure 7. TVNet compared with the baseline algorithms. Contributions to literature highlighted.

                            3.1. Results
                            3.1.1. RVNet
                                 The performance of the RVNet and the baseline algorithms are tabulated in Table 3.
                            The results show that the RVNet is better than the baseline algorithms. Results obtained
                            are shown in Figure 8.

                            Table 3. Comparative Analysis of the RVNet.

                                           Algorithms                            Average Precision                         Computing Time (ms)
                                       Tiny Yolov3 [27]                                   0.40                                             10
                                         Late Fusion                                      0.40                                             14
                                           RVNet                                          0.56                                             17

                            Figure 8. Detection result for the RVNet. The red bounding boxes and yellow circle indicate the true
                            positives and false negatives, respectively.
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                             9 of 12

                            3.1.2. TVNet
                                 The performance of the TVNet and the baseline algorithms are tabulated in
                            Tables 4 and 5. The TVNet performs the best among the different algorithms for the
                            day time subset, while reporting optimal performance for the night time subset. Results
                            are shown in Figure 9.

                            Table 4. Comparative Analysis of the TVNet on the daytime subset.

                                      Algorithms                   Average Precision              Computing Time (ms)
                                Tiny Yolov3 (Visible) [27]                0.59                            10
                               Tiny Yolov3 (Thermal) [27]                 0.58                            10
                                      Early Fusion                        0.55                            10
                                      Late Fusion                         0.56                            17
                                         TVNet                            0.61                            17

                            Table 5. Comparative Analysis of the TVNet on the nighttime subset.

                                      Algorithms                   Average Precision              Computing Time (ms)
                               Tiny Yolov3 (Visible) [27]                 0.21                            10
                                Tiny Yolov3 (Thermal)                     0.61                            10
                                     Early Fusion                         0.59                            10
                                     Late Fusion                          0.60                            20
                                        TVNet                             0.60                            17

                            Figure 9. Detection result for the TVNet. The blue bounding boxes and yellow circle indicate the
                            true positives and false negatives, respectively.
Deep Feature-Level Sensor Fusion Using Skip Connections for Real-Time Object Detection in Autonomous Driving
Electronics 2021, 10, 424                                                                                          10 of 12

                            3.1.3. Discussion
                                  Training: The RVNet and TVNet were trained using Keras on an Ubuntu 18.04
                            computer with a Nvidia Geforce 1080 GPU (NVIDIA, Santa Clara, CA, USA). The deep
                            learning network hyperparameters and learning parameters were empirically selected
                            considering the available computational resources, the real-time processing requirement
                            and training accuracy. The RVNet and the TVNet were trained with batch-size 8 and 30
                            and 120 epochs, respectively.
                                  During the network training process, the network weights associated with the different
                            epochs were periodically saved. Subsequently, we computed the testing accuracy for the
                            different network weights, and obtained the best results with 20 epochs for the RVNet
                            and 100 epochs for the TVNet. This corresponds to an early stopping-based regularization
                            reported in the literature [33].
                                  RVNet: The good performance of the RVNet can be attributed to the radar’s ability
                            to identify obstacles in varying environment, and the visible camera’s ability to delineate
                            obstacles in varying environment. The advantage of the feature level sensor fusion of radar
                            with the visible camera is clearly observed in Table 3. The RVNet reports better accuracy
                            than the visible-camera alone Tiny YOLO (baseline-1) and the late radar-camera fusion
                            method (baseline-2).
                                  Sensor Fusion: The visible camera-based baseline is affected by the inherent limitations
                            of the visible camera to low illumination and adverse weather conditions.
                                  Feature-level Fusion: In the late fusion strategy, object detection by the two sensors
                            are performed independently, without any feature-level fusion. Comparatively, the RVNet
                            performs deep learning-based sensor specific feature extraction, feature-level sensor fusion
                            and object detection within an end-to-end network. Instead of an independent object
                            detection, which are combined in the last step (baseline-2), the RVNet uses the sensor
                            specific features jointly for the object detection.
                                  TVNet: Tables 4 and 5 compare the performance of the TVNet with the baseline
                            algorithms on daytime and nighttime subsets. In the daytime subset, the TVNet reports
                            the best performance among the different algorithms.
                                  In the nighttime subset, the TVNet and the thermal camera networks report similar
                            performance. The good performance of the thermal camera-based networks is primarily
                            attributed to the performance of the thermal camera, which is invariant to the variations in
                            the environment and illumination conditions.
                                  Sensor Fusion: The advantages of feature level fusion are evident in Table 4, the
                            TVNet reports better performance than the Tiny YOLO-based visible-alone and thermal-
                            alone models.
                                  Feature-level Fusion: Comparing the sensor fusion strategies, the feature level fusion
                            strategy adopted by the TVNet is better than both the early fusion and late fusion learning
                            frameworks (Tables 4 and 5). In the early fusion learning framework, a multimodal image
                            is generated by thermal-visible image concatenation in the pre-processing step. No thermal
                            and visible camera-specific feature extraction or fusion is performed in this framework.
                                  In the late fusion learning framework, the thermal and visible camera detection
                            pipelines are independent extracting sensor specific features. However, the features are not
                            fused, and only the detection results are naively combined.
                                  Comparatively, the TVNet is an end-to-end network with simultaneous camera-
                            specific feature extraction, feature-level fusion and object detection. Consequently, the
                            TVNet reports a better detection accuracy than the baseline algorithms.

                            4. Conclusions
                                 Limitations associated with vision-based object detection are addressed using two
                            sensor fusion networks called the RVNet and the TVNet. The RVNet and TVNet are
                            proposed for visible camera-RADAR and visible-thermal camera-based object detection,
                            respectively. Compared to the literature, both the proposed networks simultaneously
                            perform sensor specific feature extraction, feature-level fusion and object detection. Both
Electronics 2021, 10, 424                                                                                                                  11 of 12

                                   the networks contain two independent feature extraction branches for sensor specific
                                   feature extraction. The output branch is a single shot detection framework which performs
                                   feature-level fusion as well as object detection. The proposed frameworks are validated on
                                   public datasets, and the results show that the RVNet and TVNet are better than baseline
                                   algorithms. The RVNet reports better accuracy than non-fusion frameworks and late fusion
                                   frameworks. Similarly, in the case of the TVNet, the TVNet reports better performance than
                                   non-fusion, early fusion and late fusion frameworks. Currently, the TVNet performance
                                   is slightly affected by the visible camera as observed in the nighttime sequences. In our
                                   future work, we will investigate methods to reduce the impact of the visible camera within
                                   the fusion framework for such challenging scenarios.

                                   Author Contributions: Conceptualization, V.J., S.M.; methodology, V.J., S.M.; software, V.J.; valida-
                                   tion, V.J.; formal analysis, V.J., S.M.; investigation, V.J., S.M.; data, V.J.; writing, V.J., S.M.; supervision,
                                   S.M. All authors have read and agreed to the published version of the manuscript.
                                   Funding: This research received no external funding.
                                   Conflicts of Interest: The authors declare no conflict of interest.

References
1.    John, V.; Karunakaran, N.M.; Guo, C.; Kidono, K.; Mita, S. Free Space, Visible and Missing Lane Marker Estimation using the
      PsiNet and Extra Trees Regression. In Proceedings of the 24th International Conference on Pattern Recognition, Beijing, China,
      20–24 August 2018; pp. 189–194.
2.    Jazayeri, A.; Cai, H.; Zheng, J.Y.; Tuceryan, M. Vehicle Detection and Tracking in Car Video Based on Motion Model. IEEE Trans.
      Intell. Transp. Syst. 2011, 12, 583–595. [CrossRef]
3.    Kafia, M.A.; Challal, Y.; Djenouria, D.; Abdelmadjid, M.D.; Badacheab, B.N. A Study of Wireless Sensor Networks for Urban
      Traffic Monitoring: Applications and Architectures. Procedia Comput. Sci. 2013, 19, 617–626. [CrossRef]
4.    Nellore, K.; Hancke, G. A Survey on Urban Traffic Management System Using Wireless Sensor Networks. Sensors 2016, 16, 157.
      [CrossRef] [PubMed]
5.    Curiac, D.-I.; Volosencu, C. Urban Traffic Control System Architecture Based on Wireless Sensor-Actuator Networks. In Pro-
      ceedings of the 2nd International Conference on Manufacturing Engineering, Quality and Production Systems, (MEQAPS’10),
      Constant, a, Romania, 3–5 September 2010.
6.    Bombini, L.; Cerri, P.; Medici, P.; Alessandretti, G. Radar-Vision Fusion for Vehicle Detection. 2006. Available online:
      http://citeseerx.ist.psu.edu/viewdoc/versions?doi=10.1.1.218.7866 (accessed on 5 January 2021).
7.    Sugimoto, S.; Tateda, H.; Takahashi, H.; Okutomi, M. Obstacle detection using millimeter-wave radar and its visualization
      on image sequence. In Proceedings of the 17th International Conference on Pattern Recognition, ICPR 2004, Cambridge, UK,
      26 August 2004; Volume 3, pp. 342–345.
8.    Fang, Y.; Masaki, I.; Horn, B. Depth-based target segmentation for intelligent vehicles: Fusion of radar and binocular stereo.
      IEEE Trans. Intell. Transp. Syst. 2002, 3, 196–202. [CrossRef]
9.    Garcia, F.; Cerri, P.; Broggi, A.; de la Escalera, A.; Armingol, J.M. Data fusion for overtaking vehicle detection based on radar and
      optical flow. In Proceedings of the IEEE Intelligent Vehicles Symposium, Alcala de Henares, Spain, 3–7 June 2012; pp. 494–499.
10.   Zhong, Z.; Liu, S.; Mathew, M.; Dubey, A. Camera Radar Fusion for Increased Reliability in ADAS Applications. Electron. Imaging
      Auton. Veh. Mach. 2018, 1, 258-1–258-4. [CrossRef]
11.   Wang, X.; Xu, L.; Sun, H.; Xin, J.; Zheng, N. On-Road Vehicle Detection and Tracking Using MMW Radar and Monovision Fusion.
      IEEE Trans. Intell. Transp. Syst. 2016, 17, 2075–2084. [CrossRef]
12.   Chadwick, S.; Maddern, W.; Newman, P. Distant Vehicle Detection Using Radar and Vision. arXiv 2019, arXiv:1901.10951.
13.   John, V.; Mita, S. RVNet: Deep Sensor Fusion of Monocular Camera and Radar for Image-based Obstacle Detection in Challenging
      Environments. In Pacific-Rim Symposium on Image and Video Technology, Proceedings of the PSIVT, Sydney, Australia, 18–22 November
      2019; Springer: Cham, Switzerland, 2019.
14.   Gaisser, F.; Jonker, P.P. Road user detection with convolutional neural networks: An application to the autonomous shuttle
      WEpod. In Proceedings of the International Conference on Machine Vision Applications (MVA), Nagoya, Japan, 8–12 May 2017;
      pp. 101–104.
15.   Milch, S.; Behrens, M. Pedestrian Detection with Radar and Computer Vision. 2001. Available online: http://citeseerx.ist.psu.ed
      u/viewdoc/versions?doi=10.1.1.20.9264 (accessed on 5 January 2021).
16.   Liu, W.; Anguelov, D.; Erhan, D.; Szegedy, C.; Reed, S.E.; Fu, C.; Berg, A.C. SSD: Single Shot MultiBox Detector. arXiv 2015,
      arXiv:1512.02325.
17.   Shah, P.; Merchant, S.N.; Desai, U.B. Multifocus and multispectral image fusion based on pixel significance using multiresolution
      decomposition. Signal Image Video Process. 2013, 7, 95–109. [CrossRef]
Electronics 2021, 10, 424                                                                                                         12 of 12

18.   Liu, Z.; Laganière, R. Context enhancement through infrared vision: A modified fusion scheme. Signal Image Video Process. 2007,
      1, 293–301. [CrossRef]
19.   Flitti, F.; Collet, C.; Slezak, E. Image fusion based on pyramidal multiband multiresolution markovian analysis. Signal Image
      Video Process. 2008, 3, 275–289. [CrossRef]
20.   Shah, P.; Reddy, B.C.S.; Merchant, S.N.; Desai, U.B. Context enhancement to reveal a camouflaged target and to assist target
      localization by fusion of multispectral surveillance videos. Signal Image Video Process. 2013, 7, 537–552. [CrossRef]
21.   Zhang, Y.; Zhang, L.; Bai, X.; Zhang, L. Infrared and visual image fusion through infrared feature extraction and visual information
      preservation. Infrared Phys. Technol. 2017, 83, 227–237. [CrossRef]
22.   John, V.; Mita, S.; Liu, Z.; Qi, B. Pedestrian detection in thermal images using adaptive fuzzy C-means clustering and convolutional
      neural networks. In Proceedings of the 14th IAPR International Conference on Machine Vision Applications (MVA), Tokyo,
      Japan, 18–22 May 2015.
23.   Shah, P.; Srikanth, T.V.; Merchant, S.N.; Desai, U.B. Multimodal image/video fusion rule using generalized pixel significance
      based on statistical properties of the neighborhood. Signal Image Video Process. 2014, 8, 723–738. [CrossRef]
24.   Shopovska, I.; Jovanov, L.; Philips, W. Deep Visible and Thermal Image Fusion for Enhanced Pedestrian Visibility. Sensors 2019,
      19, 3727. [CrossRef] [PubMed]
25.   Ma, J.; Yu, W.; Liang, P.; Li, C.; Jiang, J. FusionGAN: A generative adversarial network for infrared and visible image fusion.
      Inf. Fusion 2019, 48, 11–26. [CrossRef]
26.   Zhao, Y.; Fu, G.; Wang, H.; Zhang, S. The Fusion of Unmatched Infrared and Visible Images Based on Generative Adversarial
      Networks. Math. Probl. Eng. 2020, 2020. [CrossRef]
27.   Redmon, J.; Farhadi, A. YOLOv3: An Incremental Improvement. arXiv 2018, arXiv:1804.02767.
28.   Caesar, H.; Bankiti, V.; Lang, A.H.; Vora, S.; Liong, V.E.; Xu, Q.; Krishnan, A.; Pan, Y.; Baldan, G.; Beijbom, O. nuScenes:
      A multimodal dataset for autonomous driving. arXiv 2019, arXiv:1903.11027.
29.   FLIR. 2015. Available online: http://www.flir.com (accessed on 5 January 2021).
30.   Manjunath, A.; Liu, Y.; Henriques, B.; Engstle, A. Radar Based Object Detection and Tracking for Autonomous Driving.
      In Proceedings of the IEEE MTT-S International Conference on Microwaves for Intelligent Mobility (ICMIM), Munich, Germany,
      15–17 April 2018; pp. 1–4.
31.   John, V.; Nithilan, M.K.; Mita, S.; Tehrani, H.; Konishi, M.; Ishimaru, K.; Oishi, T. Sensor Fusion of Intensity and Depth Cues
      using the ChiNet for Semantic Segmentation of Road Scenes. In Proceedings of the Intelligent Vehicles Symposium, Changshu,
      China, 26–30 June 2018.
32.   Everingham, M.; Gool, L.; Williams, C.K.; Winn, J.; Zisserman, A. The Pascal Visual Object Classes (VOC) Challenge. Int. J.
      Comput. Vis. 2010, 88, 303–338. [CrossRef]
33.   Prechelt, L. Early Stopping-But When? In Neural Networks: Tricks of the Trade; Orr, G.B., Müller, K.R., Eds.; Springer:
      Berlin/Heidelberg, Germany, 1996; Volume 1524, pp. 55–69.
You can also read